Create Macro that runs freq if character, means if numeric

[Ian_Tfirn]

Works perfectly, thanks for your help!

[Reeza]

@heffo, use the TRIMMED option in PROC SQL to avoid the need to get rid of extra spaces issue. You should also remove the quotes from the text comparison.

[ballardw]

I would also suggest using DICTIONARY.COLUMNS in stead of SASHELP.VCOLUMN in Proc SQL. I notice very different performance, seconds versus minutes, with the number of libraries, datasets and variables involved on my system.

Since the OP says that his datasets have "hundreds of variables" this performance might be noticeable.

[Reeza]

The difference matters with the number of data sets and libraries, not columns. If that was the case I'd switch to use PROC CONTENTS or FUNCTIONS. The difference between dictionary and sashelp doesn't exist anymore the last time I tested it.

[Kurt_Bremser]

I also have the experience that @Reeza commented on. With time, SAS seems to have gotten much better at handing over where conditions through the SASHELP.Vxxxxxx views.

In earlier times, I would have the same problems with very bad performance of SASHELP:VTABLE or SASHELP.VCOLUMNS, but now (9.4 M5) there's no noticeable difference.

[heffo]

At one time or another I had some problems with missing values and quotes. So, I started to use the quotes to help this. I don't know if this is the best way to do it.
It might have been when having x ne &y or similar and not the = .

[gamotte]

Hello,

%macro freq_means(lib, mem, var);
    data _NULL_;
        set &lib..&mem.;

    if vtype(&var.)="C" then call execute('proc freq data=&lib..&mem.; table &var.; run;');
    else call execute('proc means data=&lib..&mem.; var &var.; run;');

    stop;
    run;
%mend;

%freq_means(sashelp, class, NAME);
%freq_means(sashelp, class, age);